Friends,
Titus Maccius Plautus probably isn't a name you've heard of before. But you've likely heard his most famous line. The ancient Roman playwright published Asinaria in the second century AD with the following quip:
You must spend money, if you wish to make money.
Today's artificial intelligence (AI) buildout has taken this to the extreme. Over the past three years, some of the world's most successful companies -- Google, Amazon, Microsoft, and Meta -- have taken ALL of their profits and plunged them back into building data centers. Now, some are even issuing shares and taking on debt to keep it going.
Some think this is dangerous. Some think it's wise. But last week, we got a clue for how much longer the buildout will last: at least another 17 months -- probably longer.
NVIDIA's surprise announcement
Like many companies, NVIDIA often issues guidance for the next quarter and the rest of the current year. What it does NOT usually do is issue guidance for the next fiscal year. In this case, that's the year ending on January 31, 2028 -- 17 months from now.
NVIDIA CFO Colette Kress opened last week's conference call with this line:
We expect revenue to grow by approximately 70% in fiscal 2028. This is a supply-constrained outlook.
And with that, the "AI Trade" got a whole new life.
You see, before that moment, analysts had been expecting NVIDIA to grow 44% for that year. The fact that NVIDIA upped the bullseye to 70%, and the fact that NVIDIA is often conservative in its guidance, means it wouldn't be surprising to see that figure approach 80% or more by the time January 2028 rolls around.
There's no doubt that NVIDIA is the primary beneficiary of the AI boom. The company was worth $270 billion roughly three years ago. Today, it's worth over $5 trillion. It has 20X'd in less than one presidential term!
But the big question is whether anyone else will benefit. Will Amazon, Meta, Google, and Microsoft earn a good return on their investment? What about all those piling money into Anthropic and OpenAI?
Two ways to think about how this plays out
History offers two different ways to think about this, both based on one fundamental truth: in certain industries you have to spend (a lot) of money to make (a lot) of money.
Consider James Cameron's Avatar, which debuted in 2009. The film cost an estimated $300 million+ to shoot (it would approach a half-billion dollars in 2026), one of the most expensive budgets ever at the time.
But it brought in just shy of $3 billion -- meaning it 10X'd the original investment. That's how it looks when things go right.
On the other hand, history is rife with examples of expensive buildouts that never bore fruit. Consider the supersonic airplane that is no longer in use: the Concorde. The British and French governments paid $22 billion to subsidize the plane's development. In the end:
- Only 20 Concordes were built
- Just 14 entered service
- The last five were sold for $1 apiece because there were no buyers
Will the AI buildout be Avatar-like, or Concorde-like? Our crystal ball isn't working, but we suspect it'll likely land in between. All the data we have available says demand is real and growing. But making sure the buildout matches the scale that will exist is very hard to do.
The important thing for investors to remember: while moats aren't as sexy as investing in bottlenecks, companies with moats will fare far better once supply and demand even out -- and the returns finally start coming in for the companies making huge investments.
Wishing you investing success,
|
|
Brian Feroldi, Brian Stoffel, & Brian Withers
Long-Term Mindset
|
โ